Nutriwall/Guten2.4_GGUF

Guten 2.4 is a mini lab trained story model training to create convincing human writing under 1024 tokens. It is not an instruct model and does not take instructions. Instead it attempts to continue the sentence from the prompt. It is using a llama architecture, trained on Harvard's Institutional Books 1.0, filtered down into pre-1930s text via bartholomew-dataset-v3. It is, in theory, an ethical model that does not rely on copyright infringement to train. It is a good base model for fine tuning and embedded applications, but is not finalized for a specific purpose, and could be trained further to greatly improve performance.

I would recommend changing repeat penalty to at least 2.5 (best at 5) for this model, as by default it will mainly output punctuation.